Security audit
Security checks for vulnerabilities and agentic risk
CueEdit’s account access and messaging automation match its stated purpose, with clear user approval rules and no evidence of hidden activity.
Connect only accounts you manage and approve the exact messages, targets and recipient counts before sending. CLI writes happen immediately when invoked. Auto DM rules continue until paused, so review active rules separately when disconnecting the assistant. Keep API keys out of chat and retain the default API destination unless intentionally configuring a trusted alternative.
scripts/cueedit.mjsLive publishing and destructive changes lack an enforced approval gate
A premature or unapproved tool invocation can publish real content, send real messages, activate ongoing messaging automation, or cancel a pending post. These actions are not merely previews.
Referenced artifact was not completely inspected
node scripts/cueedit.mjs accounts
Referenced artifact was not completely inspected
node scripts/cueedit.mjs posts --limit 5
Referenced artifact was not completely inspected
node scripts/cueedit.mjs comments <media_id> --unhandled
Referenced artifact was not completely inspected
` (Instagram) or `--page <id>` (Facebook) when several are connected. Run `node scripts/cueedit.mjs help` for every flag. The full endpoint list is in [referenc
Skill modifies its own code, configuration, or behavior at runtime. Self-modification enables an agent to escalate privileges, disable safety constraints, or install persistent backdoors.
**Setup:** `codex mcp add cueedit --url https://www.cueedit.io/api/mcp/mcp --bearer-token-env-var CUEEDIT_API_KEY`
**Prompt**
> Use the cueedit skill. Read-only test on my Instagram @sanskarr.tiwari: list my 3 most recent posts with comment counts, show my Auto DM rules (keyword, post, status), and for the most recent post list unhandled comments and draft a public reply for each. Do not send or change anything.
**Calls:** `cueedit/list_accounts` → `list_posts` + `list_autodm_rules` → `list_comments`
Without declared permissions the skill's intent is opaque and cannot be validated.
Although the description lists specific account-management tasks, it also explicitly triggers on broad standalone terms such as "instagram" and "reel". These can occur in informational questions or content discussions that do not request comment management, messaging, or automation.
Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.
| `send_dm` / `send_facebook_dm` | `dm <comment_id> --text "…"` / `fb-dm …` |
| `create_autodm_rule` / `list_autodm_rules` / `set_autodm_rule_status` | `rule-create --dm "…" --keywords LINK --media <id>` / `rules` / `rule-status <id> active\|paused` |
Add `--account <id>` (Instagram) or `--page <id>` (Facebook) when several are connected. Run `node scripts/cueedit.mjs help` for every flag. The full endpoint list is in [references/api.md](references/api.md). Never print, log or echo the API key, and never ask the user to paste it into chat if they can set the env var themselves.
**No account yet?** Send the user to https://www.cueedit.io/signup?ref=skill to sign up, connect Instagram (and optionally a Facebook Page), then create the key or approve the OAuth screen.
Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.
POST …/reply then POST …/dm, ~6 s apart. Stop on the first 429/402/502 and report.KEY=ce_live_…
curl -s -H "Authorization: Bearer $KEY" https://www.cueedit.io/api/v1/media?limit=5
Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.
Every call is logged to facebook_activity_log.
curl -s -H "Authorization: Bearer $KEY" "https://www.cueedit.io/api/v1/fb/posts?limit=5"
curl -s -H "Authorization: Bearer $KEY" "https://www.cueedit.io/api/v1/fb/posts/POST_ID/comments?unhandled=1"
curl -s -X POST -H "Authorization: Bearer $KEY" -H 'Content-Type: application/json' \
-d '{"text":"here is the link: https://…"}' https://www.cueedit.io/api/v1/fb/comments/COMMENT_ID/dm
Detected: suspicious.env_credential_access